Floyd C E, Tourassi G D
Department of Radiology, Duke University Medical Center, Duke University, Durham, North Carolina 27710.
Invest Radiol. 1992 Sep;27(9):667-72. doi: 10.1097/00004424-199209000-00001.
An artificial neural network (ANN) has been developed to detect nonactive circular lesions on single-slice, single-photon emission computed tomographic (SPECT) images reconstructed using filtered back projection (FBP).
The neural network is a single-layer perception which learns to identify features on the SPECT image using supervised training with a modified delta rule. The network was trained on a set of SPECT images containing clinically realistic levels of noise. The trained network was applied to a set of 120 images, and the detection performance was evaluated at several decision thresholds using receiver operating characteristic (ROC) analysis.
The trained neural network performed better than human observers for the same detection task with the same images as reflected by a significantly larger ROC curve area.
ANN can be trained successfully to perform lesion detection on reconstructed SPECT images.
已开发出一种人工神经网络（ANN），用于检测使用滤波反投影（FBP）重建的单层单光子发射计算机断层扫描（SPECT）图像上的非活性圆形病变。
该神经网络是一种单层感知器，通过使用改进的增量规则进行监督训练来学习识别SPECT图像上的特征。该网络在一组包含临床实际噪声水平的SPECT图像上进行训练。将训练后的网络应用于一组120幅图像，并使用接收器操作特征（ROC）分析在几个决策阈值下评估检测性能。
对于相同的检测任务和相同的图像，训练后的神经网络表现优于人类观察者，这体现在ROC曲线面积明显更大。
可以成功训练人工神经网络在重建的SPECT图像上执行病变检测。
